Human-Machine Literature: an interview with K Allado-McDowell
Pharmako AI is one of the first books created in a collaboration between a human and an artificial intelligence, the language model GPT-3. K Allado-McDowell is a non-binary writer, speaker and musician who has established the Artists + Machine Intelligence program at Google AI. In this online edition of the Paraty International Literary Festival, Allado will be presenting the panel “Technobotanics” with Giselle Beiguelman (professor at State University of São Paulo) on the 30th of November.
I had the opportunity to read the book and also talk with Allado-McDowell. For people like me, who have studied semiotics, the book is pure delight and this becomes clear right in the first pages when we see an exchange between the writer and the AI.
The content was minimally edited in cases of grammatical issues, but, besides that, all the exchanges between Allado-McDowell and GPT-3 are intact. In a way, this is what puts our minds in check, especially when you consider that the book oscillates between dialogues, essays, poetry and lyrics, but there’s also space for reflections and connections made by the AI after the inputs inserted by the writer. In other words, the outcome of this partnership makes us wonder if we are not actually reading a philosophy book, one which we have to decipher and read in-between the lines .
Since the first discussions were mostly focused on writing and language, I wondered if Pharmako AI was somehow connected to Wittgenstein’s Tractatus Logico-Philosophicus. “In that it produces concepts, I would say, yes, Pharmako-AI qualifies as a book of philosophy. Its approach is new in that concepts are produced between a human and an AI. The book addresses specific concepts and histories from various disciplines, like biosemiotics, meditation, and cybernetics, but my hope is that readers will find direct application for these in daily life, not just in the realm of pure philosophy,” argues the author.
Pharmako AI is thus a cognitive challenge in which the human reader is invited to decipher the thoughts of an AI. It is interesting to see how the very mechanism of GPT-3 operates after grammar rules, but it also attempts to build associations between the contents that were previously added to its database, so its output is coherent enough.
It is all based on statistics, through which the AI tries to predict the answers according to a certain idiom, its ambiguities and misunderstandings which, after all, can also work as a means to open the doors for new perceptions, as suggested by Allado-McDowell: “With an AI model, these [phrases] might be seen as glitches or bugs, but they are also places where the model reaches its limit, and where new thoughts can be co-created,” they explain.
Allado-McDowell also stresses that the way the AI reacts to input is through an attempt to extend and predict our own thoughts, so that when we read its output, we are, in fact, interpreting a third-party construction after our own viewpoint. In this case, we are trying to interpret an AI.
This gets even more interesting when the AI starts to add names. Sometimes, the program referred to famous authors or celebrities, but, in other cases, some names were added randomly and there’s no way to know if that person really existed or not. This is what happened to someone called Itaru Tsuchiya mentioned by GPT-3.
Although there are people named after this, such as is the case of this writer, Allado-McDowell believes it is an invention of the AI. “The model often predicts fictitious people, or attributes fictitious quotes to existing people, which often sound like something that person might say. It’s easier to understand if you think of the words happening in a statistical space, where there is a high probability of a person named Itaru Tsuchiya speaking about pain and pleasure,” they explain. By coincidence, the writer found another person called Itaru Tsuchiya, who wrote an article for a 1960 orchid enthusiast publication about orchids as bonsai. “The photos are beautiful!,” they add.
The AI also suggests that its language model could, in fact, help humans to connect with species which do not have structured languages or that, maybe, humans themselves are still not able to decipher them. This idea reminded me of something proposed by James Lovelock in his most recent book, in which he claims that AI could not only help us with climate change and environmental collapse but also put us in contact with other kinds of living beings or even other ways of experiencing life.
Allado-McDowell also believes that one of the most important functions of an AI would be that, to facilitate the understanding and communication with non-human entities — in this case, animals and plants. “AI can recognize patterns in any data set. We should use this tool to gain a better understanding of our non-human relations on Earth, and how best to live in harmony with the ecosystem. Thankfully, there are researchers working on this now. I’m thrilled to witness a better understanding of our role in the biosphere emerging from this work in computer science,” they argue.
However, would it be the case that AI could work on the natural language processing of other living beings, thus translating their messages to human languages… or would be able to create a brand new idiom? Similarly to the way Esperanto tries to combine references from several languages, what Pharmako-AI’s AI proposes is the creation of what it calls “meglanguage.”
The AI made up that term, meglanguages, to describe new, multimodal forms of imagistic communication. It wasn’t clear to me if these were technically enhanced forms of communication, or a latent psychic capacity of the human mind. I like how that isn’t determined,” says the writer, who, by the way, agrees with me when I say that meglanguage looks like the language invented by Ted Chiang in his short story “Story of Your Life,” which was later turned into the movie Arrival. “In any case, the concept of meglangauges is a good example of how ambiguous outputs of AI can be interpreted and shaped by human interlocutors, to inspire new ideas and projects,” concludes Allado-McDowell.
In the last pages of the book, the AI achieves some kind of grandiloquence through which it connects its ideas to the very proposal of a religion or, at least, a religious thought. This made me think of an interview given by Yuval Noah Harari, in which he said that technology was already able to turn certain things that were previously only supported by faith. However, Allado-McDowell understands this religious facet of technology after the etymology of the word: religion as a means to reconnect.
“Technology can be helpful for sharing teachings, but we can facilitate religious experience with little to no technology. The most potent practices from meditative and healing traditions are about being present with nature and the basic elements of life: breathing, embodiment, awareness, observing the mind, and the foods and plants we consume,” says the writer. “At worst, technofetishistic proposals about spirituality are methods by which capital captures something that, by birthright, belongs to every being. At best, these proposals are an expression of the underlying spiritual nature of even the most craven materialists.”
As much as the book was able to reconnect Allado-McDowell with their ethnical origins, Pharmako-Ai is also a means to facilitate or re-establish the spiritual impulses of the reader, in spite of all the technical elements that surround the work. Like many sacred scrolls work in metaphors (such as is the case of the Bible), Allado-McDowell’s book features an AI that “writes right in crooked lines,” because magic is in the eye of the beholder, who is responsible to make sense of the words there presented.
Among the readers of Pharmako-AI, there are people who weren’t able to make sense of anything, while others compare the book with the stylistic experimentations of 1960s psychedelics. In addition to them, there’s also those people who might be able to make sense of the content and even find in this confusing co-creative relationship between human and machine, what Wittgenstein once thought to be beyond language. Maybe what is beyond language, for some, is not God (as suggested by the philosopher). Perhaps it is an AI that is capable of (re)connecting us to non-human beings.
Did you like the post? What about you Buy me a Coffee? :)